IBM Think 2026: watsonx Orchestrate and Confluent redefine business operation with multi-agent AI

IBM announced on Tuesday, during the Think 2026 conference, two novelties that reinforce its bet on corporate artificial intelligence: the next generation of watsonx Orchestrate, focused on multi-agent orchestration, and IBM Confluent integration, which connects real-time data to AI models. IBM CEO Arvind Krishna pointed out that leading companies are redesigning the operation of their businesses — not just deploying more AI. The event, which brings together customers and partners in Orlando, signals a new phase in the company's strategy for the business market.

What did IBM announce in Think 2026 about multi-agent AI?

Two releases dominated the main stage. Watsonx Orchestrate, which already existed as a process automation tool, has won a multi-agent version: it is now possible to orchestrate multiple AI agents that collaborate with each other to perform complex tasks, such as approval flows, customer service and data analysis. IBM Confluent is a partnership with Confluent (data streaming platform) to allow real-time data — from sensors, logs, transactions — to directly feed watsonx AI models without the need for time-consuming ETLs.

How does watsonx Orchestrate with multiagent AI work in practice?

The watsonx Orchestrate multiagent allows companies to create “times” of specialized agents. For example: a sorting agent receives a ticket, another agent consults the inventory database, a third negotiates deadlines with a supplier, and a fourth updates the CRM system. All this is coordinated by a layer of orchestration that monitors progress and resolves conflicts. According to IBM, the tool reduces the internal call resolution time by up to 40%.

Confluent Integration: Real-time data for multiagent AI

The new IBM Confluent solves a classic bottleneck: the latency between data generation and AI decision making. Instead of processing batch data (which may have hours of delay), integration allows models to consume continuous event streams. This is vital for sectors such as finance (real-time fraud detection), logistics (dynamic delivery routing) and manufacturing (predictive maintenance). IBM claims that latency drops from minutes to milliseconds.

Redraw the business operation with multi-agent AI: what does it mean?

Arvind Krishna, CEO of IBM, was emphatic: “Companies that are winning are not just adding AI to existing processes. They are rethinking how their business works.” In practice, this means that AI is no longer an add-on — it becomes the company’s operating system. Watsonx Orchestrate and IBM Confluent are parts of this vision: AI not only performs tasks, but coordinates entire flows, from data collection to final action. IBM is betting that by 2028 80% of large companies will have multi-agent orchestration platforms in production.

Questions Common about multi-agent AI IBM Think 2026

What is IBM Confluent?

It is a joint solution of IBM with Confluent that integrates streaming data directly into watsonx AI models, eliminating the need for batch processing and allowing real-time decisions.

How can companies start using watsonx Orchestrate multiagent?

The tool is available as part of the IBM watsonx portfolio via IBM Cloud, and can be tested through early adopters programs. IBM provides consulting to map processes and create the first agents.

Does IBM support Brazilian developers?

Yes, IBM maintains innovation laboratories in São Paulo and Campinas, as well as free training programs on the IBM SkillsBuild platform. The watsonx Orchestrate documentation is available in English.

---

Original source: Financial Times - IBM Think 2026 announcement. See also the coverage on Mistral AI launches corporate platform with open models and autonomous agents to understand other business AI approaches. *